论文信息 - ART networks in automated conceptual design of structural systems

ART networks in automated conceptual design of structural systems

Abstract The present paper examines applications of adaptive resonance theory (ART) based neural networks in structural design. ART networks are vector classifiers, and their proposed use in the present work is to classify given conceptual design situations into one of several categories for which solutions are known. Once such a category is established, the actual generation of design is relegated to a procedural process associated with the category. In the present work, this approach is applied to the design of beam and frame structures for minimum weight and for maximum load carrying capacity. For such problems, the significant issues are shown to reside in selection of problem features used in the model abstraction, and in the choice of network parameters during training and classification.

[1] James A. Anderson,et al. A simple neural network generating an interactive memory , 1972 .

[2] Stephen Grossberg,et al. The ART of adaptive pattern recognition by a self-organizing neural network , 1988, Computer.

[3] D. O. Hebb,et al. The organization of behavior , 1988 .

[4] Prabhat Hajela,et al. Neural network based decomposition in optimal structural synthesis , 1991 .

[5] Prabhat Hajela,et al. Neurobiological computational models in structural analysis and design , 1991 .

[6] Teuvo Kohonen,et al. Correlation Matrix Memories , 1972, IEEE Transactions on Computers.

[7] F ROSENBLATT,et al. The perceptron: a probabilistic model for information storage and organization in the brain. , 1958, Psychological review.

[8] Stephen Grossberg,et al. A massively parallel architecture for a self-organizing neural pattern recognition machine , 1988, Comput. Vis. Graph. Image Process..